Adapter Fine Tuning

AbstractLarge language models are shown to present privacy risks through memorization of training data, andseveral recent works have studied such risks for the pre-training phase. Little attention, however, has been …...